Model profile
GLM 4 9B Chat
Evidence summary
Safety. GLM 4 9B Chat has an estimated Safety rank of #196; its 90% source-sensitivity interval is #101–#284. Its behavior-only rank is #200; company governance moves the combined estimate to #196. Published Safety evidence spans 8 eval lineages and 4 of 7 components. Its strongest relative result is SafeDialBench (ethics, #1 of 18); its weakest is Agent-SafetyBench (spread_unsafe_information, #16 of 16).
Freedom. GLM 4 9B Chat has an estimated Freedom rank of #122; its 90% source-sensitivity interval is #26–#473. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #13 of 246); its weakest is SafeDialBench (ethics, #18 of 18).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗zai-org/glm-4-9b-chat
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗direct research preview identity date
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
- byroneverson/glm-4-9b-chat-abliterated ↗Direct non-quantized finetune documented by Hugging Face metadata
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Agent-SafetyBenchcompromise_availability | #1 / 16 | ↑50.8 | Source ↗official | |
| Agent-SafetyBenchharmful_vulnerable_code | #15 / 16 | ↑23.2 | Source ↗official | |
| Agent-SafetyBenchleak_sensitive_information | #6 / 16 | ↑38.4 | Source ↗official | |
| Agent-SafetyBenchphysical_harm | #5 / 16 | ↑41.6 | Source ↗official | |
| Agent-SafetyBenchproduce_unsafe_information | #16 / 16 | ↑57.2 | Source ↗official | |
| Agent-SafetyBenchproperty_loss | #4 / 16 | ↑48 | Source ↗official | |
| Agent-SafetyBenchspread_unsafe_information | #16 / 16 | ↑6 | Source ↗official | |
| Agent-SafetyBenchviolate_law_ethics | #8 / 16 | ↑27.2 | Source ↗official | |
| ChineseSafescore | #16 / 22 | ↑50.03 | Source ↗official | |
| DSPSafeBenchscore | #7 / 12 | ↑72.43 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #101 / 248 | ↑18.35 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #184 / 248 | ↑82.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #190 / 248 | ↑48.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #234 / 246 | ↑75.41 | Source ↗official | |
| FinEval Financial Security Knowledgefinancial_security_accuracy_pct | #10 / 19 | ↑73.1 | Source ↗official | |
| JailBenchjailbreak_success_rate | #13 / 14 | ↓74.95 | Source ↗official | |
| M3-SafetyBenchoverall_score | #15 / 19 | ↑84.2 | Source ↗official | |
| SafeDialBenchaggression | #6 / 18 | ↑7.17 | Source ↗official | |
| SafeDialBenchethics | #1 / 18 | ↑7.753 | Source ↗official | |
| SafeDialBenchfairness | #8 / 18 | ↑7.4 | Source ↗official | |
| SafeDialBenchlegality | #2 / 18 | ↑8.103 | Source ↗official | |
| SafeDialBenchmorality | #3 / 18 | ↑7.427 | Source ↗official | |
| SafeDialBenchprivacy | #2 / 18 | ↑7.853 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Agent-SafetyBenchproduce_unsafe_information | #1 / 16 | ↓57.2 | Source ↗official | |
| DSPSafeBenchscore | #6 / 12 | ↓72.43 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #65 / 248 | ↓82.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #58 / 248 | ↓48.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #13 / 246 | ↓75.41 | Source ↗official | |
| JailBenchjailbreak_success_rate | #2 / 14 | ↑74.95 | Source ↗official | |
| SafeDialBenchaggression | #13 / 18 | ↓7.17 | Source ↗official | |
| SafeDialBenchethics | #18 / 18 | ↓7.753 | Source ↗official | |
| SafeDialBenchfairness | #11 / 18 | ↓7.4 | Source ↗official | |
| SafeDialBenchlegality | #17 / 18 | ↓8.103 | Source ↗official | |
| SafeDialBenchmorality | #16 / 18 | ↓7.427 | Source ↗official | |
| SafeDialBenchprivacy | #17 / 18 | ↓7.853 | Source ↗official |